Test-retest reliability of arterial spin labeling with common labeling strategies

Abstract

Purpose: To compare the test-retest reproducibility of three variants of arterial spin labeling (ASL): pseudo-continuous (pCASL), pulsed (PASL) and continuous (CASL).

Materials and methods: Twelve healthy subjects were scanned on a 3.0T scanner with PASL, CASL, and pCASL. Scans were repeated within-session, after 1 hour, and after 1 week to assess reproducibility at different scan intervals.

Results: Comparison of within-subject coefficients of variation (wsCV) demonstrated high within-session reproducibility (ie, low wsCV) for CASL-based methods (gray matter [GM] wsCV for pCASL: 3.5% ± 0.02%, CASL: 4.1% ± 0.07%) compared to PASL (wsCV: 7.5% ± 0.06%), due to the higher signal-to-noise ratio (SNR) associated with continuous labeling, evident in the 20% gain in temporal SNR and 58% gain in raw SNR for pCASL relative to PASL. At the 1-week scan interval, comparable reproducibility between PASL (GM wsCV 9.2% ± 0.12%) and pCASL (GM wsCV 8.5% ± 0.14%) was observed, indicating the dominance of physiological fluctuations.

Conclusion: Although all three approaches are capable of measuring cerebral blood flow within a few minutes of scanning, the high precision and SNR of pCASL, with its insensitivity to vessel geometry, make it an appealing method for future ASL application studies.

Copyright © 2011 Wiley-Liss, Inc.

Figures

Figure 1

CBF maps (in units of ml/100g/min) spatially normalized to MNI space calculated from one session (40 averages) of each of the three sequences (from top to bottom: PASL, pCASL, CASL) for 4 representative subjects.

Figure 2

Plots of GM (left) and WM (right) CBF values for the different scan intervals of the three sequences. Black line represents the line of equality.

Figure 3

Temporal SNR (tSNR) and raw SNR (rSNR) of pCASL and CASL, normalized to PASL. Error bars represent SD.

Figure 4

Absolute GM signal differences between perfusion images for each additional average, averaged across all subjects. SDs are omitted for better visualization of the difference between the three sequences. Lines represent results from fitting a 1/√NA function to the data.

Figure 5

Box plot of male and female GM CBF values for the three sequences. For each group, the thick line inside the box represents the median, and the two ends of the box are the 1st and 3rd quartile. The T-bars for each box extend to 1.5 times the height of the box. Values that lie outside the T-bars are represented by circles and identified as outliers.

Figure 6

Sample CBF maps in ml/100g/min from all 7 subjects obtained using CASL. Notice two of the maps appear to have limited signal in portions of the posterior circulation (arrows).